---
title: "Not so sexually modern after all:"
subtitle: "Homonegativity and prejudice against open and age-gap relationships"
author:
  - name: "*Blinded authors*"
        
shorttitle: "Not So Modern After All"
shortauthor: "Blinded authors"
date: today
date-format: "D MMMM YYYY"
abstract: |
  How widespread is sexual liberalism in tolerant societies? Theoretical and descriptive evidence suggest an overall liberalization in societal views on topics such as women's rights and homosexuality. Yet, relying on sexual norms theory, this study unveils persistent sexual prejudice. We differentiate between "normalized" sexual issues, like same-gender marriage, which have gained mainstream acceptance, and "non-normalized" issues, such as non-traditional sexual practices and relationships, which remain stigmatized. Through a conjoint experiment in Catalonia, Spain, we investigate public attitudes towards adoptive parents with varying sexual orientations, relationship types, and age differences, confirming that discriminatory preferences are prevalent in contexts with low social desirability. By highlighting the continued prejudice against both normative and non-normative sexual issues, this research contributes to our understanding of the dynamics of sexual attitudes and the challenges facing LGBTQ+ politics and rights.
format:
  simple-article-pdf:
    keep-tex: true
    indent: false
    classoption: 
     - 11pt
  simple-article-html: default
  wordcount-pdf:
    toc: true
    fig-align: center
    cap-location: margin
reference-section-title: References
bibliography: bibliography.bib
---

## Introduction

Recent scholarship suggests that in modern democracies a majority of citizens have become sexually modern [@spierings2017; @lancaster2019]. This thesis is congruent with @inglehart1977's prediction, which anticipated that increasing prosperity in Western societies would shift the focus from survival-based concerns to cultural politics and foster more liberal attitudes. Empirical evidence from surveys supports this, revealing that today's citizens exhibit more open views on a variety of sociocultural issues than in the past. Key areas of liberalization include women's rights [@lancaster2022], the liberalization of attitudes towards homosexuality [@caughey2019], and the acceptance of single, co-habiting, and same-gender parenthood [@vera-toscano2021]. This liberalization process, scholars argue, has also reached to traditionally conservative groups, giving rise "sexually-modern nativists" – that is, individuals who simultaneously embrace positive views for homosexuals and a less traditional role of women in society, but hold negative attitudes towards immigration [@spierings2017; @lancaster2019].

The most popular explanations for this shift have been benevolent. @inglehart1977 argued that conservative attitudes would diminish through the process of cohort replacement, as younger, post-materialist individuals gained influence in politics; something that has recently found confirming evidence [@lancaster2022; @ekstam2023]. For the acceptance of sexual and gender minorities, explanations have also included the liberalising effect of the enactment of inclusive laws [@abou-chadi2019], as well as increased personal interactions with sexual and gender minorities [@ayoub2017; @garretson2018].

However, recently, researchers have provided a more critical perspective on the rapid liberalization of societal attitudes, suggesting that citizens' expressed liberal attitudes may not be as genuine or unconditional as previously assumed [@turnbull-dugarte2023]. According to this thesis, many citizens may endorse liberal values, not out of conviction, but primarily because they perceive these values as integral to Western national identity [@laegaard2007]. The discrepancy between expressed and privately held attitudes does not stem solely from conscious concealment, as early social desirability research suggested [@edwards1953]. Rather, recent research on implicit attitudes by @lodge2013 and @pérez2016 shows that prejudiced reactions also occur without conscious awareness. This combination of conscious and unconscious processes in the expression of prosocial attitudes, together with the motivation to differentiate from ethnic out-groups, generates a social norm that outwardly supports sexually progressive values, leading individuals to express support for LGBTQ+ rights despite privately holding opposing views [@turnbull-dugarte2023]. This 'sexual norm veil' would contribute to explain why a vast majority of citizens in most Western European countries supports the rights of gays and lesbians [@gubbala2023; @wilson2020], yet continue to show discriminatory preferences against them in contexts with low social desirability such as experiments [@turnbull-dugarte2022; @magni2021]. However, this theory does not fully account for why while support for same-gender marriage and adoption has become mainstream, negative attitudes towards other sexual issues, such as LGBTQ+ symbols, are more openly expressed [@lopezortega].

Also, students of sexual modern attitudes have so far disregarded the increasing relationship diversity that modern societies experience. A recent study in the UK found that a third of heterosexual men were open to having more than one spouse or long-term partner, along with 11% of women [@thomas2024]. Additional studies have found that one in five people surveyed in the US [@haupert2017] and Canada [@fairbrother2019] have experience with non-monogamy. Additionally, more women are marrying younger men than ever before [@lehmiller2011] and absolute age differences seem to be steadily increasing in countries such as Sweden [@kolk2015]. In a context in which far-right leaders and religious conservative groups make common cause against increasing relationship diversity [@aqeedi2023], it is crucial to incorporate attitudes towards relationship diversity into the scope of sexual modernism.

In this research note, we introduce novel theoretical reasoning and empirical evidence that elucidates the contrasting levels of support for different sexual issues, challenging the perceived extent of sexual liberalism in Western societies. Drawing from sociological insights [@doan2014], our analysis differentiates between *normalized* and *non-normalized sexual issues*. *Normalized sexual issues* include those that have been mainstreamed in modern democratic contexts, typically revolving around formal rights such as same-gender marriage and adoption. The enactment of inclusive legislation [@abou-chadi2019], alongside the politicization of these issues by politicians, and the advocacy by civil rights activists and educators [@harrison2017] have cultivated a supportive social norm towards same-gender couples and families. Consequently, there is a heightened social cost for citizens to express opposition to LGB couples and their parenting rights.

Underlying homonegativity and prejudice, however, may persist despite this social climate. Research has experimentally documented the continued existence of homonegativity against gay parents [@turnbull-dugarte2022]. This may be linked to the prevailing cultural belief that children require both a mother and a father to develop healthily, despite a lack of empirical evidence supporting this view [@biblarz2010; @patterson2017]. Importantly, the nature of objections varies based on the family structure, with more positive attitudes typically directed towards two-mother households than two-father families. This may reflect cultural assumptions of men being less nurturing than women, and fatherhood being a more peripheral aspect of masculine identity compared to motherhood for women \[ @fiske2002 @biblarz2010\]. Furthermore, the proliferation of campaigns perpetuating "grooming" tropes against LGBTQ+ people, but more specifically gay men [@cassisa2024], serve as a breeding ground for prejudice against queer parents in relation to childhood. All of this suggests that, even when there is a social norm for tolerance and acceptance of normalized rights such as LGBTQ+ parenting, especially in contexts where the law allows it for a long time, under reduced social costs, negativity towards LGBTQ+ parents, especially gay men, may manifest.

In contrast, the movement to normalize formal rights for LGB individuals has not extended equally to other sexual and gender minorities, nor has it fully embraced support for *non-normalized sexual issues* [@murib2023; @yi2013]. Some authors argue that selective inclusion of certain aspects of LGBTQ+ rights has resulted in less progress for the freedoms of less privileged queer populations facing more blatant discrimination [@murib2018; @murib2023], and lower public acceptance for non-normalized sexual realities, such as non-monogamous relationships [@yi2013], but arguably also non-traditional sexual practices, and age-gap relationships. This argument is supported by evidence that shows persistent open prejudice against couples who differ in age [@banks2001; @collisson2020] and that engage in non-monogamy [@kaufman2022], and leads us to expect not only that prejudice against both age-gap and non-monogamous couples is present, but that is also greater than the prejudice against LGB parents.

That said, the prejudice associated with normalized and non-normalized sexual issues often intersects. Literature on respectability politics suggests that LGB couples enjoy broader societal support as long as they adhere to heteronormative family norms [@strolovitch2018]. Violating these norms results in heightened prejudice compared to their peers. The evidence for this respectability penalty is mixed in the domain of selecting politicians [@everitt2021; @jones2022]. However, research indicates that matters related to children are distinct. Owing to the sexual stigma associated with queer individuals [@dottisani2020; @herek1986], education and parenting are areas where society likely holds queer people to higher standards. Thus, while LGB partnership and parenting have been normalized, we suggest that when LGB individuals, especially gay parents, engage in non-normalized sexual relationship types they experience compounded discrimination.

On top of exploring attitudes towards diverse parenting, this paper seeks to understand whether two of the most prevalent traditional correlates of negativity towards LGBTQ+ issues, i.e. (right-wing) ideology and (older) age (for a review, @adamczyk2019), are also associated to more conservative views around the topics of relationship diversity.

To investigate these dynamics, we employ a pre-registered[^1] conjoint experiment focusing on sexual parenting rights conducted in Catalonia, Spain---a region at the forefront of liberal attitudes according to existing data--. This experiment investigates respondents' sexual attitudes in favor of hypothetical scenarios where participants choose between potential adoptive parents differing in traits such as sexual orientation, relationship type (monogamous or open), and age difference. These traits vary in their level of social normativity, allowing us to assess attitudes towards normative versus non-normative sexual issues in a context of reduced social desirability bias.

[^1]: For an overview of the pre-analysis plan, visit: <https://osf.io/gjn8v/?view_only=9b5377fec6ba4a2a890da2d1ffd31ae8>.

We believe our study makes three key contributions. First, by showing that in contexts with low social desirability, individuals exhibit more homonegative attitudes than they might express in public opinion surveys, especially when it has to do with two-men parenting. Second, we show that this veiled homonegativity affects citizens across the ideological spectrum, even if left-wing citizens hold more positive views for two-women couples. Third, we show that these homonegative attitudes, while significant, are small relative to the still prevailing conservative views towards non-normative parental characteristics. In sum, this study holds important implications for students of sexual attitudes, LGBTQ+ politics, and the liberalization phenomenon more generally.

## Context

To contextualize sexually modern attitudes in Catalonia, we draw on data from the European Social Survey (ESS) to compare the region's tolerance towards gays and lesbians with that of other European countries. Additionally, we use data from the Centro de Investigaciones Sociológicas (Centre of Sociological Research, CIS) to examine support for polyamorous and open couples.

The ESS has included various questions regarding lesbians, gays, and homosexual couples in all their survey waves. Starting in Round 1, the survey featured the question "Gay men and lesbians should be free to live their own life as they wish," which has been present in Rounds 1 through 10. Later rounds also introduced questions such as "Gay male and lesbian couples should have the same rights to adopt children as straight couples" and "If a close family member was a gay man or a lesbian, I would feel ashamed", both of which appeared in Rounds 8 through 10.

Historically, Catalonia has demonstrated a high level of tolerance towards gays and lesbians living as they wish. In Round 1, 78% of the population agreed or strongly agreed with this statement, rising to 88% in Round 10. Additionally, there is widespread rejection of the idea of being ashamed of having a gay or lesbian family member, with over 80% of Catalans disagreeing with this statement in Round 10. Regarding attitudes towards adoption by homosexual couples, nearly 80% of the Catalan population supports this right, a figure comparable to that in other socially progressive countries such as the United Kingdom, France, Norway, and Sweden.

In contrast, attitudes towards polyamorous and open couples are more divided. According to the 2023 CIS survey on post-pandemic social and emotional relationships, 55% of Catalans believe that a person can have two or more affective-sexual relationships simultaneously. However, less than half (45%) believe that partners can agree to have sex with others outside the relationship without being romantically involved.

Given these findings, one might expect Catalonia—a region that strongly supports the freedom of homosexuals to live as they wish, rejects the stigma of having a homosexual family member, and broadly endorses the right of homosexual couples to adopt children—to be an unlikely case for homonegative values. However, the divided opinion on polyamorous and open relationships suggests potential prejudice against these types of relationships when considering adoption applications.

## Experiment

The data for this experiment was drawn from the 2022 Sociopolitical Survey, published by the Centre d'Estudis d'Opinió (Centre of Opinion Studies, CEO), a part of the Regional Government of Catalonia. The sample was obtained during the first wave of the Panel Ciutadà (Citizen's Panel) and comprised 5,569 respondents, of whom 4,298 completed the survey online, while 1,271 responded on paper. For the purposes of this study, only the online portion of the sample was used, as the conjoint experiment was conducted digitally. The survey was available in both Spanish and Catalan.

Given the potential for social desirability bias in questions concerning the rights of homosexual couples in Spain—where both marriage and adoption enjoy majority support after years of legal recognition—alternative methods are necessary to assess whether homophobia persists in society and if it intensifies when prejudices are compounded by factors such as non-monogamous relationships or significant age differences between partners. Due to the sensitivity of these issues, a conjoint experiment is a suitable approach for uncovering latent negative attitudes towards sexually diverse couples. As @bansak2021 notes, "when respondents evaluate several attributes simultaneously, they may be less concerned that researchers will connect their choices to one specific attribute," thus mitigating the social desirability effect.

The conjoint experiment involved a hypothetical adoption scenario in which respondents were asked to choose between two couples after evaluating several characteristics. These characteristics included the sexual orientation of the partners (inferred from their names' gender combinations), the type of relationship (closed or open), the age of the partners, their education level, their family values (strict or permissive), and the reason for the adoption request. The latter three characteristics were included to prevent participants from discerning the true focus of the experiment. The experiment featured 21 different couples: ten heterosexual couples, six lesbian couples, and five gay male couples.

Respondents were presented with a table (Table 1) showing both couples and were required to select either 'Couple 1' or 'Couple 2' after evaluating the provided information. Each respondent completed this task five times. Importantly, participation was voluntary; respondents could choose to skip a task or the entire experiment if they wished. Table 1 illustrates an example of two random couples that respondents could encounter during the experiment.

| Characteristics | Couple 1 | Couple 2 |
|-------------------|--------------------------|----------------------------|
| Names (sexual orientation) | Marc and Joan | Jose and Marta |
| Type of relationship | Open couple, have relationships with other people | Closed couple, only have relationships with each other |
| Age | 23 and 53 | 32 and 34 |
| Education level | Universitary studies | Universitary studies |
| Values | Strict | Permissive |
| Reason to adopt | Can't have biological kids | Can't have biological kids |

: Example of conjoint task

## Results

To analyze the results of the conjoint experiment, we utilize Marginal Means (MM), given the methodological challenges associated with Average Marginal Component Effects (AMCE), particularly when selecting a reference category, as discussed by @leeper2020. Marginal Means are calculated as the mean outcome of a conjoint feature level across all instances presented to respondents, averaged across the other features. In the figures, the null effect is represented by a line at 0.5, while the effects of the different features are depicted by 95% confidence intervals on the x-axis. If the Marginal Means values are below 0.5, the feature's effect is negative; if they are above 0.5, the effect is positive. The further the interval is from 0.5, the greater the effect of the characteristic.

We will first present the results for all MM values across all features, followed by the interactions between these features and other categorical variables. Although the experiment presented only the partners' names without explicitly indicating their sexual orientation, we use gender as a cue to refer to heterosexual (man and woman couple), lesbian (two-women couple), and gay (two-men couple) pairs. These labels typically correspond to these couple compositions, although we acknowledge that this is not universally the case.

As shown in Figure 1, the experiment reveals a significant preference among respondents for heterosexual couples (0.51) over LGB couples (0.49) when selecting potential parents for adoption. This finding suggests that even in a 'socially modern' society like Catalonia, discrimination against sexual minorities persists, despite widespread tolerance for LGB individuals and support for adoption by homosexual couples in surveys.

However, when distinguishing between gay and lesbian couples, the bias against lesbian couples diminishes (0.5), while the prejudice against gay male couples is both significant and negative (0.48). This indicates that the negative effect observed for non-heterosexual couples compared to heterosexual couples is primarily driven by the bias against gay male couples. Figure 2, which provides a pairwise comparison of the Marginal Means for each type of couple, confirms these conclusions: while heterosexual couples are marginally preferred over lesbian couples, both heterosexual and lesbian couples are significantly favored over gay male couples by respondents when selecting adoption parent candidates.

```{r plot 1, echo=FALSE, fig.height=6, fig.width=7, message=FALSE, warning=FALSE, fig.cap = "Marginal means for the general model"}

library(tidyverse)
library(cregg)
library(patchwork)

conjoint <- read_csv("conjoint_data.csv")
conjoint <- conjoint |> mutate(across(where(is.character), as.factor))
conjoint$sexuality_couple <- factor(conjoint$sexuality_couple,
                                    levels = c("Man and woman couple", "Two-women couple", "Two-men couple"))

f1 <- choice ~ sexuality_couple + relationship_couple + age_couple + education_couple + values_couple + reason_couple + pairwise_so_couple
mm <- mm(conjoint, f1, id = ~order, weigths = ~weights)

mm |> 
  mutate(feature = case_when(
    feature == "pairwise_so_couple" ~ "Couple's sexuality",
    feature == "sexuality_couple" ~ "Couple's sexuality by sex of partners",
    feature == "relationship_couple" ~ "Type of relationship",
    feature == "age_couple" ~ "Couple's age",
    feature == "education_couple" ~ "Couple's education",
    feature == "values_couple" ~ "Couple's family values",
    feature == "reason_couple" ~ "Reason to adopt",),
    feature = factor(feature,
                   levels = c("Couple's sexuality",
                              "Couple's sexuality by sex of partners",
                              "Type of relationship",
                              "Couple's age",
                              "Couple's education",
                              "Couple's family values",
                              "Reason to adopt"))) |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      group = level,
                      color = feature,
                      xmin = lower, 
                      xmax = upper),
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
  geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_x = if_else(mm$estimate > 0.5,0.03,-0.03)) + 
  scale_y_discrete(limits = rev) +
  scale_color_manual(values = c("#DB69A7",
                                "#D91122",
                                "#D56F40",
                                "#FFBC42",
                                "#218380",
                                "#3397E4",
                                "#79137D")) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3","0.4","0.5","0.6","0.7")) +
  theme_minimal() +
  theme(legend.position = "none",
        panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm"),
        strip.placement = "outside",
        panel.spacing.y = unit(0, "lines")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(feature),
             scales = "free_y",
             ncol = 1,
             strip.position = "left",
             labeller = label_wrap_gen(width=15))
```

```{r plot 2, echo=FALSE, fig.height=3, fig.width=7, message=FALSE, warning=FALSE, fig.cap = "Pairwise test for Marginal Mean differences."}
pairw_3_est <- mm |> 
  filter(feature == "sexuality_couple") |> 
  mutate(level = case_when(
    level == "Man and woman couple" ~ "man_woman",
    level == "Two-women couple" ~ "two_women",
    level == "Two-men couple" ~ "two_men")) |> 
  select(estimate, level) |> 
  pivot_wider(names_from = level, values_from = estimate) |> 
  mutate(diff_gay_het = two_men - man_woman,
         diff_les_het = two_women - man_woman,
         diff_gay_les = two_men - two_women) |> 
  select(diff_gay_het, diff_les_het, diff_gay_les) |> 
  pivot_longer(everything(), names_to = "level", values_to = "estimate") 

pairw_3_low <- mm |> 
  filter(feature == "sexuality_couple") |> 
  mutate(level = case_when(
    level == "Man and woman couple" ~ "man_woman",
    level == "Two-women couple" ~ "two_women",
    level == "Two-men couple" ~ "two_men")) |> 
  select(lower, level) |> 
  pivot_wider(names_from = level, values_from = lower) |> 
  mutate(diff_gay_het = two_men - man_woman,
         diff_les_het = two_women - man_woman,
         diff_gay_les = two_men - two_women) |> 
  select(diff_gay_het, diff_les_het, diff_gay_les) |> 
  pivot_longer(everything(), names_to = "level", values_to = "lower") 

pairw_3_up <- mm |> 
  filter(feature == "sexuality_couple") |> 
  mutate(level = case_when(
    level == "Man and woman couple" ~ "man_woman",
    level == "Two-women couple" ~ "two_women",
    level == "Two-men couple" ~ "two_men")) |> 
  select(upper, level) |> 
  pivot_wider(names_from = level, values_from = upper) |> 
  mutate(diff_gay_het = two_men - man_woman,
         diff_les_het = two_women - man_woman,
         diff_gay_les = two_men - two_women) |> 
  select(diff_gay_het, diff_les_het, diff_gay_les) |> 
  pivot_longer(everything(), names_to = "level", values_to = "upper") 

pairw_3 <- left_join(pairw_3_est, pairw_3_low)

pairw_3 <- left_join(pairw_3, pairw_3_up)

pairw_3 |> 
     mutate(level = case_when(
     level == "diff_gay_het" ~ "Two-men vs Man and woman",
     level == "diff_les_het" ~ "Two-women vs Man and woman",
     level == "diff_gay_les" ~ "Two-men vs Two-women"),
     level = factor(level,
                    levels = c("Two-men vs Man and woman",
                               "Two-women vs Man and woman",
                               "Two-men vs Two-women"))) |> 
  ggplot() +
  geom_vline(xintercept = 0, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      group = level,
                      color = level,
                      xmin = lower, 
                      xmax = upper),
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
  geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_color_manual(values = c("#D91122",
                                "#D91122",
                                "#D91122")) +
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(-0.05,0.05),
                      breaks = c(-0.05,-0.025,0.0,0.025,0.05),
                      labels = c("-0.05","-0.025","0.0","0.025","0.05")) +
  theme_minimal() +
  theme(legend.position = "none",
        panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm"),
        strip.placement = "outside",
        panel.spacing.y = unit(0, "lines")) +
  labs(y = "", 
       x = "Marginal means differences",
       title = "") 
```

Figure 1 also illustrates that non-traditional relationships, such as open couples and those with a significant age difference between partners, experience even greater prejudice from respondents, exhibiting more pronounced negative effects than those associated with the couple's sexuality. This is particularly evident in the case of relationship type, where the negative effect for open couples is the second strongest in the model (0.4), indicating that non-monogamous couples face significant discrimination when it comes to family matters.

Regarding age differences in couples, those with less of an age gap are significantly more preferred (0.53) than those with a larger age difference (0.47). One might argue that this effect varies depending on whether the older partner in heterosexual relationships is the man or the woman. However, although relationships where the man is considerably older are more penalized than those where the woman is older, respondents generally penalize both types of relationships in comparison to those with a smaller age gap (see Figure A2 in the Appendix).

As described in the experiment, several 'filler' variables were included to prevent respondents from discerning the true focus of the research and providing socially desirable responses. Two of these variables are particularly noteworthy: education level and the reason for adoption. With regards to education level, results show a significant preference for couples with higher education (0.54) over those with lower education levels (0.51 for secondary education and 0.45 for primary education), which is a clear indication of the class bias in the context of adoption eligibility.

Conversely, the reason for adoption emerges as the most influential factor for respondents when selecting a couple, as shown in Figure 1. The strongest negative effect in the entire model is observed for couples who already have biological children and wish to adopt (0.36). When this characteristic is combined with the couple's sexuality (see Figure A3 in the Appendix), the general hierarchy of preferences persists for couples who cannot have biological children, with heterosexual couples being preferred over lesbian couples, followed by gay couples. However, for couples who already have biological children and wish to adopt, the penalty is generally higher, particularly for gay couples.

When examining the interaction between non-traditional relationship attributes and the sexuality of the couples, we find that the type of relationship does not produce distinct preferences but rather exhibits a compound effect between these factors (see Figure 3). The negative effect towards lesbian and gay couples is amplified when they are non-monogamous, although the preference hierarchy—heterosexual, lesbian, and gay couples—remains intact. Differences in marginal means reveal a slightly more negative (albeit not significant) effect against open gay couples. Overall, these results reflect the compounded nature of prejudice, with particularly low preferences for lesbian (0.4) and gay couples that engage in open relationships (0.38).

Regarding the couple's age gap, Figure 4 demonstrates that the interaction influences preferences for non-heterosexual couples, albeit differently for gay and lesbian couples. While heterosexual and lesbian couples with a smaller age difference are significantly more preferred than gay couples with similar characteristics, both gay and lesbian couples with a larger age difference are significantly less preferred than heterosexual couples. Differences in marginal means indicate that the prejudice against couples with a larger age gap is slightly more pronounced for lesbian couples, although this difference is only marginally greater than the prejudice faced by gay and age-gap heterosexual couples. This interaction highlights how existing prejudice against couples with a significant age difference is even more pronounced for gay and lesbian couples.

```{r plot 3, echo=FALSE, fig.height=3, fig.width=7, message=FALSE, warning=FALSE, fig.cap = "Interaction between sexuality of the couple and type of relationship."}
mm_h1_1 <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ relationship_couple)
diff_h1_1 <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm_differences", by = ~ relationship_couple)

mm_h1_1_plot <- mm_h1_1 |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      group = BY,
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
    geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3","0.4","0.5","0.6","0.7")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY))

diff_h1_1_plot <- diff_h1_1 |> 
  ggplot() +
  geom_vline(xintercept = 0, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      group = BY,
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
    geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.25) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(-0.3,0.3),
                     breaks = c(-0.2,0,0.2),
                     labels = c("-0.2","0","0.2")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        axis.text.y = element_blank(),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Differences in \nmarginal means",
       title = "") +
  facet_wrap(vars(BY), 
             labeller = label_wrap_gen(width = 25, multi_line = TRUE))

mm_h1_1_plot + diff_h1_1_plot +
  plot_layout(widths = c(5, 2.5))
```

```{r plot 4, echo=FALSE, fig.height=3, fig.width=7, message=FALSE, warning=FALSE, fig.cap = "Interaction between sexuality of the couple and age difference of partners"}
mm_h1_2 <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ age_couple)
diff_h1_2 <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm_differences", by = ~ age_couple)

mm_h1_2_plot <- mm_h1_2 |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
      geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3","0.4","0.5","0.6","0.7")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY))

diff_h1_2_plot <- diff_h1_2 |> 
  ggplot() +
  geom_vline(xintercept = 0, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      group = BY,
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
    geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.25) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(-0.3,0.3),
                     breaks = c(-0.2,0,0.2),
                     labels = c("-0.2","0","0.2")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        axis.text.y = element_blank(),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Differences in \nmarginal means",
       title = "") +
  facet_wrap(vars(BY), 
             labeller = label_wrap_gen(width = 30, multi_line = TRUE))

mm_h1_2_plot + diff_h1_2_plot+
  plot_layout(widths = c(5, 2.5))
```

As mentioned in the previous section, we also want to examine how ideology affects attitudes towards diverse parenting. Figure 5 shows that centrist and right-wing respondents significantly prefer heterosexual couples to gay and lesbian couples in hypothetical an adoption process. However, gay couples are the most discriminated against in this interaction, as they show a negative effect across all ideologies. Nonetheless, this effect diminishes in left-wing respondents, who also prefer lesbian couples more than both heterosexual and gay couples.

```{r plot 5, echo=FALSE, fig.height=3, fig.width=7, message=FALSE, warning=FALSE, fig.cap = "Interaction between sexuality of the couple and ideology of respondent"}
mm_h1_3 <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ ideology)

mm_h1_3 |> 
  mutate(BY = case_when(
    BY == "Left" ~ "Left (0-4)",
    BY == "Centre" ~ "Centre (5)",
    BY == "Right" ~ "Right (6-10)"),
    BY = factor(BY,
                     levels = c("Left (0-4)",
                                "Centre (5)",
                                "Right (6-10)"))) |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
      geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3", "0.4","0.5","0.6","0.7")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY))
```

Regarding the interaction between the sexuality of the couple and the age of the respondent, Figure A4, in the Appendix, shows that heterosexual couples are more preferred than homosexual couples as the age of respondents increases, with only the two youngest cohorts showing no significant preferences for these couples. On the other hand, gay couples are significantly less preferred in most age groups except the youngest (18-24), while lesbian couples do not show significant estimates in most cohorts, with a significant negative effect only among respondents aged 50-64.

These results confirm that in 'modern' societies, age and ideology still structure tolerance of homosexual couples and support for their rights, although not completely. However, this is not the case for non-traditional relationships, as neither age nor ideology turns the views for open couples or couples with larger age differences into positive, showing that prejudice against these types of couples is present in all social groups and is not directly connected to generational effect (Figures A5 and A6 in Appendix).

## Conclusion

This study confirms the expectation that while people continue to hold prejudice against LGB individuals in the context of parenting, this prejudice is small compared to the negativity experienced by individuals engaged in age-gap and non-monogamous relationships. This finding strengthens previous research showing that, even in modern societies, persistent social negativity exists towards LGBTQ+ people under low social costs [@turnbull-dugarte2023; @magni2021], particularly in the realm of parenting [@turnbull-dugarte2022]. Given that adoption procedures are highly competitive - in Spain, applications far exceed available children [@observatoriodelainfancia] - even small prejudices can systematically tilt decisions against certain candidates. This is especially consequential for LGB couples who are restricted to national adoptions due to international barriers [@prats2020]. In such competitive contexts, where agencies must differentiate between similarly qualified candidates, subtle biases can accumulate across multiple evaluation points, converting statistically modest preferences into meaningful barriers to adoption.

Regarding whether there is heightened prejudice against LGB individuals involved in non-normative relationships, our results show that there is not an increased penalty but rather an additive effect. The prejudice against LGB couples is compounded, rather than multiplied, when they engage in non-normative relationships. In practical terms, this translates to greater overall prejudice against LGB couples because studies indicate that LGB individuals are more likely to engage in less conventional types of relationships, as evidenced by greater age differences between partners [@boyd2003; @lehmiller2011] and a higher incidence of non-monogamous relationships [@haupert2017]. Thus, even if LGB individuals in non-normative relationships are not facing double-layered discrimination, they still encounter more prejudice than their heterosexual counterparts simply because they are more likely to engage in non-normative sexual behaviors and relationships.

However, it is important to note that our sample may not fully represent the actual population of couples seeking to adopt children, which is a limitation of our study [@cuesta2023]. Despite this, our results offer valuable insights into the broader societal biases against same-gender couples.

Overall, this research underscores the importance of studying overlooked and less normalized attitudes within the field of sexual modernism for two main reasons. First, the preference structure for these attitudes appears to differ substantially from that of other sexual attitudes. Previous research has demonstrated that sexual attitudes vary depending on whether they concern abstract values or concrete policies [@dottisani2020], formal rights or informal privileges [@doan2014], or involve symbolic politics [@lopezortega]. While traditional predictors such as ideology and age are applicable to newer forms of symbolic prejudice [@lopezortega], prevalent negative attitudes towards non-normative relationship types seem to be unrelated to age and politically unaligned.

Second, we show that attitudes towards non-normative relationship types continue to be the target of particular prejudice. This is significant, especially as society evolves and new forms of relationships become more common, with more people forming identities around these types of relationship preferences [@barker2005; @klesse2014]. Consequently, these relationships should inevitably become a greater focus in the ever-expanding research on sexual modern politics. Further research could also help mitigate the external validity issues of this article—speaking to the actual population of couples seeking to adopt children [@cuesta2023], and better understanding how the conclusions of this paper translate to other contexts with varying levels of sexual modernism.

### Funding Statement

This research is based in data collected by the Centre d'Estudis d'Opinió (CEO), a public institution that embedded the experiment and items within their survey [*Enquesta sociopolítica. 2022*](https://ceo.gencat.cat/ca/estudis/registre-estudis-dopinio/estudis-dopinio-ceo/societat/detall/index.html?id=8708 "Enquesta sociopolítica. 2022")*.*

### Competing Interests

None.

## References

::: {#refs}
:::

## Appendix {.appendix}

\setcounter{table}{0}
\renewcommand{\thetable}{A\arabic{table}}
\setcounter{figure}{0}
\renewcommand{\thefigure}{A\arabic{figure}}

**Table A1:** Comparison between weithgted sample demographics and official data in Catalonia (in %)

|   |   | Weighted sample | Official data (Institut d’Estadística de Catalunya, 2022) |
|------------------|------------------|------------------|-------------------|
| Sex | Female population. | 51 | 51 |
|  | Male population. | 49 | 49 |
| Age | 16-24 years old | 11 | 11 |
|  | 25-34 years old | 14 | 14 |
|  | 35-49 years old | 28 | 28 |
|  | 50-64 years old | 24 | 24 |
|  | 64+ years old | 23 | 23 |
| Education level | Primary education | 17 | 17 |
|  | Secondary education | 51 | 51 |
|  | Higher education | 32 | 32 |

```{r table-3, echo=FALSE, message=FALSE, warning=FALSE}
#| tbl-cap: Marginal Means estimates

library(flextable)

flextable(mm |> 
  select(-outcome,
         -statistic) |>
  mutate(estimate = round(estimate, digits = 2),
         std.error = round(std.error, digits = 3),
         z = round(z, digits = 2),
         p = round(p, digits = 2),
         lower = round(lower, digits = 2),
         upper = round(upper, digits = 2),
    feature = case_when(
    feature == "pairwise_so_couple" ~ "Couple's sexuality",
    feature == "sexuality_couple" ~ "Couple's sexuality by sex of partners",
    feature == "relationship_couple" ~ "Type of relationship",
    feature == "age_couple" ~ "Couple's age",
    feature == "education_couple" ~ "Couple's education",
    feature == "values_couple" ~ "Family values",
    feature == "reason_couple" ~ "Reason to adopt",),
    feature = factor(feature,
                   levels = c("Couple's sexuality",
                              "Couple's sexuality by sex of partners",
                              "Type of relationship",
                              "Couple's age",
                              "Couple's education",
                              "Family values",
                              "Reason to adopt")))) |> 
  border_inner_h(part="all") |> 
  merge_v(j = "feature") |> 
  valign(j = "feature", 
         valign = "top") 

```

**Figure S1:** Frequencies of conjoint features

```{r plot s1, echo=FALSE, fig.height=6, fig.width=10, message=FALSE, warning=FALSE}
freq <- cj_freqs(conjoint, f1, id = ~order)

freq |> 
  mutate(feature = case_when(
    feature == "pairwise_so_couple" ~ "Couple's sexuality",
    feature == "sexuality_couple" ~ "Couple's sexuality by sex of partners",
    feature == "relationship_couple" ~ "Type of relationship",
    feature == "age_couple" ~ "Couple's age",
    feature == "education_couple" ~ "Couple's education",
    feature == "values_couple" ~ "Couple's family values",
    feature == "reason_couple" ~ "Reason to adopt",),
    feature = factor(feature,
                   levels = c("Couple's sexuality",
                              "Couple's sexuality by sex of partners",
                              "Type of relationship",
                              "Couple's age",
                              "Couple's education",
                              "Couple's family values",
                              "Reason to adopt"))) |> 
  ggplot() +
  geom_col(aes(y = level,
                      x = estimate, 
                      group = level,
                      fill = feature)) + 
  scale_y_discrete(limits = rev) +
  scale_fill_manual(values = c("#DB69A7",
                               "#D91122",
                               "#D56F40",
                               "#FFBC42",
                               "#218380",
                               "#3397E4",
                               "#79137D")) +
  theme_minimal() +
  theme(legend.position = "none",
        panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 13),
        plot.margin = margin(0,0.5,0.5,0, "cm"),
        strip.placement = "outside",
        panel.spacing.y = unit(0, "lines")) +
  labs(y = "", 
       x = "Frequencies",
       title = "") +
  facet_wrap(vars(feature),
             scales = "free_y",
             ncol = 1,
             strip.position = "left",
             labeller = label_wrap_gen(width=15))
```

**Figure A1:** Interaction between gender of the oldest partner and age difference of partners

```{r plot a1, echo=FALSE, fig.height=3, fig.width=7, message=FALSE, warning=FALSE}
mm_age <- cj(conjoint, choice ~ agediff_sex_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ age_couple)
diff_age <- cj(conjoint, choice ~ agediff_sex_couple, id = ~order, weights = ~weights, estimate = "mm_difference", by = ~ age_couple)

mm_age_plot <- mm_age |> 
    mutate(level = case_when(
    level == "Man has more age than woman" ~ "Man has more\nage than woman",
    level == "Woman has more age than man" ~ "Woman has more\nage than man")) |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
      geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
   scale_x_continuous(limits = c(0.4,0.6),
                     breaks = c(0.4,0.5,0.6),
                     labels = c("0.4","0.5","0.6")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY))

diff_age_plot <- diff_age |> 
  ggplot() +
  geom_vline(xintercept = 0, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      group = BY,
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
    geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(-0.2,0.2),
                     breaks = c(-0.2,0,0.2),
                     labels = c("-0.2","0","0.2")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        axis.text.y = element_blank(),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Differences in \nmarginal means",
       title = "") +
  facet_wrap(vars(BY), 
             labeller = label_wrap_gen(width = 30, multi_line = TRUE))

mm_age_plot + diff_age_plot +
  plot_layout(widths = c(5, 2.5))
```

**Figure A2:** Interaction Interaction between sexuality of the couple and reason to adopt

```{r plot a2, echo=FALSE, fig.height=5, fig.width=7, message=FALSE, warning=FALSE}
mm_reason <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ reason_couple)
diff_reason <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm_difference", by = ~ reason_couple)

mm_reason_plot <- mm_reason |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
      geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3","0.4","0.5","0.6","0.7")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY), 
             labeller = label_wrap_gen(width = 40, multi_line = TRUE))

diff_reason_plot <- diff_reason |> 
  ggplot() +
  geom_vline(xintercept = 0, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      group = BY,
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
    geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(-0.3,0.3),
                     breaks = c(-0.3,0,0.3),
                     labels = c("-0.3","0","0.3")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Differences in marginal means",
       title = "") +
  facet_wrap(vars(BY), 
             labeller = label_wrap_gen(width = 90, multi_line = TRUE))

mm_reason_plot / diff_reason_plot
```

**Figure A3:** Interaction between sexuality of the couple and age of respondents

```{r plot a3, echo=FALSE, fig.height=4, fig.width=7, message=FALSE, warning=FALSE}
mm_h1_4 <- cj(conjoint, choice ~ sexuality_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ age)

mm_h1_4 |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
      geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.3) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3","0.4","0.5","0.6","0.7")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY))
```

**Figure A4:** Interaction between type of relationship and age of respondents

```{r plot a4, echo=FALSE, fig.height=4, fig.width=7, message=FALSE, warning=FALSE}
mm_test_1 <- cj(conjoint, choice ~ relationship_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ age)

mm_test_1 |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
      geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3","0.4","0.5","0.6","0.7")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY))
```

**Figure A5:** Interaction between age difference of partners and age of respondents

```{r plot a5, echo=FALSE, fig.height=4, fig.width=7, message=FALSE, warning=FALSE}
mm_test_2 <- cj(conjoint, choice ~ age_couple, id = ~order, weights = ~weights, estimate = "mm", by = ~ age)

mm_test_2 |> 
  ggplot() +
  geom_vline(xintercept = 0.5, 
             size = 0.3,
             linetype = "dashed") +
  geom_pointrange(aes(y = level,
                      x = estimate, 
                      xmin = lower, 
                      xmax = upper),
                  color = "#D91122",
                  size = 0.2,
                  position = position_dodge(width = 1)) + 
      geom_label(aes(y = level,
                 x = estimate,
                 label = round(estimate, digits = 2),
                 group = level),
             size = 2.75,
             nudge_y = 0.2) + 
  scale_y_discrete(limits = rev) +
  scale_x_continuous(limits = c(0.3,0.7),
                     breaks = c(0.3,0.4,0.5,0.6,0.7),
                     labels = c("0.3","0.4","0.5","0.6","0.7")) +
  theme_light() +
  theme(panel.grid.minor = element_blank(),
        plot.background = element_rect(color = "white"),
        text = element_text(size = 11),
        plot.margin = margin(0,0.5,0.5,0, "cm")) +
  labs(y = "", 
       x = "Marginal means",
       title = "") +
  facet_wrap(vars(BY))
```
